AI tools for Compute

ERNIE 5.5 Flash Base is Baidu's released foundation checkpoint for further model training. The sparse language model has 280 billion total parameters and activates approximately

P-JiT 21K is Linum's released research checkpoint for text-to-image generation. The two-billion-parameter model was trained on 660 million samples to investigate what a smaller

RelayVSR performs 4x super-resolution on real-world video. It uses a large generative model selectively to recover detail and a smaller network for the rest of the sequence.

Spark-X2.6-1.7B is a compact open-weight language model designed for practical on-device AI. It supports coding, planning, tool use, reasoning, and multilingual tasks while keep

GPT-6.1 Terra is OpenAI's model for efficient visual reasoning, computer use, and written work. It accepts text and images and generates text, supporting applications that inter

MiniMax M3.1 Flash is a language model optimized for complex agent tasks, coding, computer use, and cybersecurity workflows. It improves the provider's Flash serving recipe to r

Open Duet is a self-hostable AI coworker application for named agents with persistent instructions and memory. Each coworker can use an isolated computer, follow recurring routi

Olmo-core 3 is Ai2's open training infrastructure for large language models and mixture-of-experts systems. A redesigned MoE training stack distributes model development across

OpenAI Dots are persistent agents that work on delegated projects using their own cloud computer and connected applications. Powered by GPT-6 Astra, they carry context across co

GPT-6.1 Sol is OpenAI's model for coding, computer use, and professional work. It improves on GPT-6 Sol across agentic tasks while reducing cached-input cost for applications th

GPT-6 Sol and Luna are OpenAI model tiers for coding, professional work, and computer-use agents. They bring methods from GPT-6 Astra into faster, lower-cost models, with prompt

AlphaGenome Atlas is a searchable resource for investigating how DNA changes may affect molecular biology. It precomputes AlphaGenome predictions for every possible single-nucle

Nex-N2.5 is a model family for sustained agent work across computers, browsers, coding tasks, and knowledge workflows. Its Mini and Pro models combine vision with action-oriente

Marigold V2 converts a single image into detailed depth and other dense visual predictions. Its model family covers ordinary depth, geometry behind transparent surfaces, surface

GPT-6 Astra is OpenAI model for coding, research, computer interaction, and professional work. It can produce documents, spreadsheets, presentations, and applications.

Claude Opus 5 is Anthropic's Opus-tier model for long-running agents, coding, and professional knowledge work. It is positioned as a major upgrade over prior Opus models, with s

GenCeption is a research model that repurposes a pretrained video generation model into a unified feed-forward vision learner. Instead of training a separate model head for each

GNM, the Generative aNthropometric Model and Ecosystem from Google, is an open ecosystem for parametric human models and related perception stacks. The repository starts with GN

Brain2Qwerty is Meta's non-invasive brain-to-text research system for decoding typed language from brain activity. Brain2Qwerty v2 is described as an end-to-end pipeline that ap